Seasonal Movement Patterns and Summertime Use of Thermal Refuge Areas by Muskellunge in the Nontidal Potomac River, Maryland

Native to the St. Lawrence–Great Lakes and Mississippi River basins, the Muskellunge Esox masquinongy is considered a coolwater fish (Fuller et al. 2020), with an upper temperature tolerance of 30°C (Scott and Crossman 1973). Many native Muskellunge populations were lost during the 20th century due to a combination of overharvest, habitat degradation, pollution, and liberal bag limits (Kerr 2011). Muskellunge populations rebounded as water quality standards improved, angler ethics began favoring catch and release, and management priorities changed. This led to subsequent range expansions (via introductions) into several mid-Atlantic and southeastern states (Kerr 2011). Despite the new science that is rapidly evolving around southern Muskellunge populations, multiple data gaps still exist and information is needed to make informed management decisions (Crane et al. 2015).

Understanding life history characteristics of fish species is imperative prior to implementing appropriate management strategies (Winemiller 2005). Managers typically rely on creel and length limits to maintain desired size structures (Oele et al. 2016), ensure sustainable recruitment (Parkinson et al. 2004), and balance ecological interactions with competing fishes and other aquatic biota (Carpenter et al. 1985). However, size and creel limits must consider growth rates, seasonal life history patterns, angling vulnerability, and mortality to be effective (Woodward and Griffin 2003). These factors become more important for larger and longer-lived species, as limited knowledge of spatial life histories can result in unsatisfactory management outcomes over longer temporal scales (Smith et al. 2018). This is especially true for managing trophy fisheries that require 5–10 years or more to evaluate success. The ability to identify seasonal movement patterns, key spawning areas, and refuge areas is necessary to protect life histories and sustain long-lived fishes, such as sturgeon (Auer 1996) and Muskellunge (Dombeck et al. 1984; Zorn et al. 1998).

Muskellunge were introduced into the nontidal Potomac River—most likely via angler introduction—during the late 1980s. A self-sustaining, naturally reproducing population developed, and the Potomac River population remains Maryland's only Muskellunge fishery. Muskellunge have been documented along approximately 241 km of river, with the highest densities occurring in a 129-km stretch of river within Washington County, Maryland (MDDNR 2017). Mid-Atlantic and southeastern rivers frequently reach temperatures that are considered stressful (30°C) to Muskellunge (Scott and Crossman 1973). The successful establishment of Muskellunge fisheries in these systems has resulted in the need to better understand population dynamics in fisheries where water temperatures often exceed the species' thermal threshold (Crane et al. 2015). To date, few studies have been conducted on Muskellunge movement patterns during stressful warm periods. However, work from Virginia documented high angler usage on the James River (Bauerlien et al. 2021) and New River (Brenden et al. 2007; Doss et al. 2019) populations.

Crane et al. (2015) identified Muskellunge populations in the southern part of the species' range as being greatly understudied and the most threatened due to temperature increases in a changing climate. Beck and Brooks (2003) observed adult Muskellunge selecting deeper, cooler water when surface temperatures reached 25°C, despite lower dissolved oxygen (DO) concentrations (<3.0 mg/L) during summer months in a southern Illinois reservoir. In a Tennessee reservoir, Cole and Bettoli (2014) reported a thermal range for adult Muskellunge of 22.3°C and concluded that fish never occupied water temperatures greater than 27.5°C and rarely occupied water warmer than 25°C, despite the sufficient availability of DO.

The use of radiotelemetry in understanding seasonal movement patterns is critical to identify and protect spawning locations (Crossman 1990; Younk et al. 1996; Jennings et al. 2011; Morrison and Warren 2015), determine habitat types and preferences (Brenden et al. 2006), and identify seasonal vulnerabilities to angling (Cole and Bettoli 2014). Multiple radiotelemetry studies have investigated Muskellunge behaviors (i.e., seasonal movements and habitat preferences, spawning behavior and homing) in riverine environments (Minor and Crossman 1978; Younk et al. 1996; Brenden et al. 2006), but limited work has been done on populations in the southern part of the Muskellunge's range. The objectives of this study were to (1) understand seasonal Muskellunge movements and summertime habitat preferences in a southern riverine environment and (2) document Muskellunge use of thermal refugia during summer months.

METHODS

Study area

The freshwater (nontidal) Potomac River has a 29,900-km² watershed that is shared by Pennsylvania, Maryland, West Virginia, Virginia, and the District of Columbia. Largely unregulated by dams, this river supports a natural flow regime (Hitt et al. 2020), with only two major barriers (hydroelectric dams: Dam 4 and Dam 5) in the freshwater system (Figure 1). The Chesapeake and Ohio (C&O) Canal parallels the freshwater portion of the Potomac River for 296 km between Cumberland (Maryland) and Georgetown (District of Columbia), offering shoreline access to the river along much of the Maryland shore (Figure 1).

Fish collection and tagging

Muskellunge were collected from the river and tracked in two river sections separated by dams. Study area 1 was located upstream of Williamsport, Maryland, and between two hydroelectric dams (Dam 5 and the Potomac Edison Roller Dam) that limit fish migration; thus, study area 1 emulates a closed population (Figure 1). Study area 2 was located downstream of Dam 4 and represented a more open system with no significant barrier to downstream movement (Figure 1). During March 2017, 14 Muskellunge (7 males and 7 females) were collected from study area 1 using pulsed-DC boat electrofishing. Individual fish were measured (mm), weighed (kg), and assigned a sex by visually examining the urogenital pore (Lebeau and Pageau 1989) prior to surgery (Table 1). Fish were anesthetized using electronarcosis (Hudson et al. 2011) until stage 4 anesthesia was achieved (Summerfelt and Smith 1990) and then were surgically implanted with an Advanced Telemetry Systems (ATS) F1170 Series radio transmitter (ATS, Isanti, Minnesota) operating on a unique frequency (pulse rate = 30 pulses/min) within the 48.0–51.9-MHz range and employing an internal antenna. Additionally, each fish was externally marked with two yellow dart tags (type PDL; Hallprint, Hindmarsh Valley, South Australia) anchored into the basal pterygiophores proximal to the dorsal fin on the fish's left side. All surgical equipment and radio transmitters were sanitized using a 2.6% glutaraldehyde solution for at least 24 h prior to surgery and were rinsed with sterile saline water. Each transmitter was checked for functionality prior to insertion by activating the tag and holding it next to the receiver. Nonabsorbable synthetic monofilament sutures (3/0) in conjunction with a reverse cutting needle were used to close the incision using the simple interrupted suture pattern described by Wagner et al. (2010).

TABLE 1. Data describing individual Muskellunge from the Potomac River radiotelemetry study (2017–2020).

Year tagged	TL (mm)	Weight (g)	Sex	Transmitter number	Study area	Number of detections	Period tracked
							Start date	End date
2017	908	5.05	M	450	1	10	Mar 12, 2017	Jul 3, 2017
2017	921	5.52	F	830	1	10	Apr 20, 2017	Sep 5, 2017
2017	883	4.79	M	710	1	15	Mar 12, 2017	Sep 21, 2017
2017	1,010	7.12	F	950	1	15	Mar 12, 2017	Nov 15, 2017
2017	927	5.3	M	670a	1	6	Mar 12, 2017	Apr 26, 2017
2017	1,130	11.04	F	911	1	18	Mar 12, 2017	Nov 15, 2017
2017	1,086	9.65	F	870b	1	12	Apr 3, 2017	Aug 3, 2017
2017	1,061	8.62	F	469	1	10	Mar 12, 2017	Aug 1, 2017
2017	895	4.63	M	750	1	19	Mar 12, 2017	Nov 15, 2017
2017	953	5.22	M	840	1	15	Mar 23, 2017	Nov 15, 2017
2017	965	4.28	F	770	1	16	Mar 12, 2017	Nov 15, 2017
2017	813	3.42	M	930	1	15	Mar 23, 2017	Nov 15, 2017
2017	959	5.06	F	680	1	10	Mar 23, 2017	Nov 15, 2017
2018	787	3.26	M	410	1	12	Mar 23, 2017	Nov 15, 2017
2018	1,003	5.96	F	201	1	63	Apr 11, 2018	Aug 31, 2020
2018	876	4.54	M	171a	1	25	Apr 11, 2018	Mar 28, 2019
2018	1,162	12.5	F	341	1	61	Apr 11, 2018	Aug 12, 2020
2018	711	2.38	M	281a	1	24	Apr 11, 2018	Apr 17, 2019
2018	908	4.85	M	111a	1	4	Apr 11, 2018	May 24, 2018
2018	756	2.9	M	301a	1, 2	17	Apr 11, 2018	Sep 4, 2019
2018	864	4.46	F	291a	1	11	Apr 11, 2018	Jul 11, 2019
2018	1,010	6.41	F	32	1	41	Apr 11, 2018	Apr 7, 2020
2018	978	6.79	M	352a	1	5	Apr 11, 2018	Jun 19, 2018
2018	768	2.91	M	71	1	60	Apr 11, 2018	Aug 31, 2020
2018	800	3.8	F	12	1	60	Apr 12, 2018	Aug 31, 2020
2018	826	3.7	F	52	2	16	Apr 12, 2018	May 5, 2020
2018	984	7.2	M	362	2	45	Apr 12, 2018	Jan 23, 2020
2018	991	6.57	M	221	2	45	Apr 12, 2018	Aug 31, 2020
2018	787	3.02	M	91	2	54	Apr 12, 2018	Aug 31, 2020
2018	781	2.86	M	251	2	49	Apr 12, 2018	Aug 31, 2020
2018	984	7.13	M	191a	2	36	Apr 12, 2018	Sep 4, 2019
2018	1,086	10.91	F	271	2	59	Apr 12, 2018	Aug 31, 2020
2018	1,035	8.72	F	152	2	62	Apr 12, 2018	Aug 31, 2020
2018	927	5.11	M	131a	2	2	Apr 12, 2018	Jun 21, 2018
2018	826	3.92	F	241a	2	21	Apr 12, 2018	May 22, 2019
2018	737	2.7	F	262a	2	49	Apr 12, 2018	May 5, 2020

• ^a Mortality or tag loss/malfunction.
• ^b Angler harvest.

During March 8–28, 2018, we collected 22 Muskellunge from study areas 1 and 2. Eleven fish (6 males and 5 females) were collected from each study area using the previously described methods. All data collection, tagging, and postoperative procedures were consistent with the methods previously described, with one notable exception: fish received surgically implanted ATS F1850 Series radio transmitters. The transmitters weighed 25 g out of water, and the transmitter burden (transmitter weight: fish weight ratio) was kept below 2%. Each transmitter operated on a unique frequency (pulse rate = 35 pulses/min) within the 48.0–51.9-MHz range and included a 40-cm external antenna and a specialized mortality sensor that omitted double the pulse rate (70 pulses/min) when movements ceased for a 24-h period. Any movement would subsequently turn the mortality function off, and the tag would operate normally again. To accommodate the external antenna, modified surgical procedures were used; a 4-cm incision was made proximate to the ventral midline and anterior to the pelvic girdle (linea alba) in the peritoneal cavity; using the scalpel handle to shield the viscera, a 16-gauge hypodermic needle was inserted posterior to the incision to allow the antenna wire to pass through the body (Wagner et al. 2010). Transmitters were then inserted lengthwise into the peritoneal cavity. Nonabsorbable synthetic monofilament sutures (3/0) in conjunction with a reverse cutting needle were used to close the incision using the simple interrupted suture pattern described by Wagner et al. (2010).

Tracking

Tracking was conducted weekly throughout the summer months and at least once per month during other seasons (fall, winter, and spring) from March 2017 through August 2020. Specific fish locations were determined by homing with a R2000 VHF receiver equipped with a tunable, adjustable ground loop antenna (ATS). Once it was determined that the proximity to the fish was within 20 m, the antenna was removed and the fish was located using the open end of the coaxial cable (Brenden et al. 2006) to determine a more accurate location (within 5 m). In 2017, all tracking efforts were conducted by boat due to the limited range of detection for the F1170 Series transmitters. In 2018–2020, tracking was primarily conducted by vehicle due to the proximity of the C&O Canal towpath, which paralleled the river throughout the study site. River width and the total distance of the C&O Canal towpath to the river's edge in the study area(s) did not exceed the predetermined range of the F1850 Series radio transmitters. Vehicle tracking maximized efficiency for covering a larger geographic area during each tracking event. A linear position was recorded from the tracker's location on the canal towpath except during summer months, when more precise homing to the fish's location occurred to determine presence or absence in refuge habitats (obvious tributaries, springs, or groundwater seeps). During refuge occupancy, the tracker would navigate, either on foot or by boat, to within 5 m of the fish's actual location and a signal could be detected using just the open end of the coaxial cable.

Telemetry data analysis

Data were categorized by season: spring (March–May), summer (June–August), fall (September–November), and winter (December–February). All radiotelemetry locations were imported into ArcMap 10.2 (ESRI, Redlands, California) to quantify fish movements. Linear distance traveled between tracking events was measured for each individual fish and divided by the number of days at large between locating events. These data were expressed as meters per day and averaged by season to represent a seasonal movement rate. Seasonal movement data were averaged for each individual and then across individuals for each season. One-way ANOVA was used to determine differences among seasons. Tukey–Kramer post hoc analyses were used to interpret results computed from one-way ANOVA. Relative movements were calculated based on individual fish position in relation to the postsurgical release point (designated as river kilometer “0”). Upstream movements were represented as a positive integer, while downstream movements were negative values. Additionally, the distance to each thermal refuge area (TRA; described below) from the postsurgical release point was determined following the same upstream/downstream criteria described above. These data were used only to illustrate spatial migration (upstream versus downstream) and homing behaviors of Muskellunge. Mortality signals from the F1850 Series transmitters were detected occasionally from 2018 through 2020. Attempts to recover these samples were unsuccessful; therefore, we deemed it reasonable to treat these signals as live samples until the signal and lack of upstream movements were consistently detected for greater than 2 months. Once the mortality signals and lack of upstream movement were consistent for over 2 months, we stopped tracking these fish and the data collected during the 2 months were excluded from analyses.

Suspected TRAs during summer tracking were identified by obvious springs, upwellings, or tributaries that entered the main-stem Potomac River and were known to be habitually occupied by Muskellunge during summer months (J.H., personal observations). Thermal refuge areas were confirmed if summer temperatures were less than main-stem river temperatures, as described below.

Temperature and discharge

HOBO Onset Pro V2 temperature loggers were deployed in the main-stem river (n = 2) and suspected TRAs (n = 9) within the study area from June through September in 2017–2020. Six temperature loggers were calibrated, deployed, and retrieved in accordance with procedures outlined by the Maryland Department of Natural Resources (MDDNR 2016). Three additional temperature loggers were deployed by the U.S. Geological Survey (USGS) in 2021 following the procedures described by Snyder et al. (2015). River flow data were downloaded from USGS gauge 01613000 at Hancock, Maryland (USGS 2020). Both temperature and flow statistics were calculated, and daily and weekly mean, minimum, and maximum values were used as independent, fixed effects for TRA occupancy modeling.

Temperature data analysis

Temperature data from the HOBO loggers deployed within the main-stem river were summarized by daily and weekly averages (June–August 2017–2020). Temperature data collected from eight TRAs during this study were compared to a main-stem temperature logger located near Dam 5 (Figure 1) to determine relative differences. The second main-stem logger, located adjacent to the “West Virginia Musky Hole,” was deployed on the Maryland shore during 2020 and compared to that TRA only. A one-way ANOVA was conducted to test for mean monthly temperature differences for each of the nine TRAs in June, July, and August relative to the main-stem Potomac River. Tukey's honestly significant difference test was performed to identify pairwise differences for each TRA relative to main-stem river temperatures.

Thermal refuge area occupancy modeling

Detections (i.e., presence) of Muskellunge within a TRA (e.g., creek mouth, spring, or groundwater upwelling) were modeled as the dependent variable using a binomial generalized linear mixed model (GLMM) from the lme4 package in R (Bates et al. 2015; R Core Team 2021). Detections were pooled across all 4 years (2017–2020) because we were interested in temperature- and flow-specific effects on refuge occupancy irrespective of year. Detections were designated as “1” for presence and “0” for absence and were used as the response variable in the models. A suite of GLMMs was constructed for both daily mean and weekly mean flow (cubic meters per second) and temperature (°C) variables along with fish TL (mm) and sex as fixed effects, while individual fish (i.e., tag identification number) were treated as the random effect. Model fit was evaluated using the DHARMa package (Hartig 2020) to test for overdispersion and distribution of fitted versus observed residuals for each model. Model selection was based on the lowest Bayesian information criterion (BIC) score (Schwartz 1978) and model weight. The top model was illustrated using a box-and-whisker plot of the odds ratios from model predictions in 1°C intervals.

RESULTS

After tracking of initially tagged Muskellunge for 3 months, the retention of surgically implanted tags was estimated at 94%. Two fish that received tags in 2018 exhibited a lack of movements and consistent mortality signals 2 and 3 months after surgery. One fish from the 2017 sample only provided a signal for 1 month posttagging. We were unable to determine whether these fish had shed their tags or were deceased; therefore, these individuals were excluded from analyses. The F1170 Series radio transmitters used in 2017 provided reliable data for 9 months (March–November). Number of detections averaged 18.7 per fish and ranged from 10 to 26 per fish (Table 1). Three of the 13 fish were lost during summer 2017, with one of these fish harvested by an angler during TRA occupancy.

Of the 22 fish that were initially equipped with F1850 Series radio transmitters, 9 individuals still had functional transmitters and were located by the end of the study (Table 1). One additional fish experienced tag loss or mortality in 2018. Six fish produced consistent mortality signals and no movement during 2019. An additional four fish were removed from the study during 2020. Fish 301, a 755-mm male, was initially tagged in study area 1 (2018) and was undetected from April 2018 until the following spring (March 2019), when it was located in study area 2. This individual traveled over two dams and 36.7 river kilometers downstream, representing the furthest documented migration in this study.

Seasonal Movements

Movements were greatest during transitional periods throughout the spring as fish moved from deeper wintering habitats to more shallow spawning areas (P < 0.01; Figure 2). Apparent spawning site fidelity was observed throughout this study; 91% of the Muskellunge in this sample returned to the same location(s) during consecutive spawning seasons. The mean distance traveled by Muskellunge that exhibited this homing behavior was 6,126 m (SE = 1,207) and ranged from 290 to 25,000 m. During winter, fall, and summer, Muskellunge were more sedentary, with mean movement rates of 43.98 (SE = 9.98), 49.18 (SE = 16.79), and 41.64 (SE = 6.93) m/d, respectively (Figure 2).

Thermal Refuge Area Temperature Analysis

Temperature data were collected from nine candidate TRA sites (creek mouths, seeps, springs, etc.) with known summer occupancy by Muskellunge. Two additional suspected TRAs were identified during this study based on Muskellunge occupancy patterns in summer months; the suspected TRAs had characteristics similar to those of the nine TRAs we monitored for temperature and were assumed to have similar thermal benefits. However, no temperature data were collected for these sites. The percentages of total Muskellunge detections at TRAs by season (2017–2020) were 69% (422 detections) in summer, 14% (21 detections) in the fall, 5% (4 detections) in winter, and 2% (5 detections) in the spring. Differences in mean temperature from each TRA relative to the main-stem Potomac River ranged from −14.4°C at the Gift Road Spring site during July to −0.32°C at the West Virginia Musky Hole in June. One-way ANOVAs indicated significant differences in mean temperature among all nine TRAs compared to the main stem (P < 0.01). Pairwise comparisons using Tukey's honestly significant difference test indicated significantly lower mean monthly temperatures except for the West Virginia Musky Hole during June (Table 2). Relative movement rates of all fish with active tags at the conclusion of this study (n = 9) are illustrated in Figure 3 with respect to TRA proximity. Thermal refuge area homing was consistent throughout the study (Figure 3), with 92% of the fish returning to the same refugia during consecutive years. Among the 13 fish tagged in 2017, 10 individuals were detected in the same TRA during summer 2018 as they were located during 2017, but these data were excluded from analyses because continuous tracking of these fish ceased in November 2017.

TABLE 2. Results of pairwise comparisons (Tukey's honestly significant difference test) for June, July, and August from nine thermal refuge areas where temperature data were collected (RT = river temp; CON = Conococheague Creek; LCON = Little Conococheague Creek; SJR = Saint James Run; GRS = Gift Road Spring; SS = Shepherd Spring; RR = Rattlesnake Run; RMR = Rockymarsh Run; TR = Town Run; WVH = West Virginia Hole [“Musky Hole”]). Difference in temperature (°C) relative to the main-stem Potomac River (ΔTemp) is presented with lower and upper confidence limits (95% LCL and UCL).

Comparison	Month	ΔTemp	95% LCL	95% UCL	P-value
RT-CON	Jun	−0.88	−0.64	−1.14	0.004
RT-CON	Jul	−0.95	−0.68	−1.23	0.001
RT-CON	Aug	−1.0	−0.76	−1.22	0.0006
RT-LCON	Jun	−3.61	−3.37	−3.85	0.00
RT-LCON	Jul	−4.56	−4.31	−4.82	0.00
RT-LCON	Aug	−3.95	−3.68	−4.24	0.00
RT-SJR	Jun	−4.11	−3.85	−4.39	0.00
RT-SJR	Jul	−5.62	−5.35	−5.92	0.00
RT-SJR	Aug	−4.76	−4.48	−5.05	0.00
RT-GRS	Jun	−11.5	−10.5	−12.6	0.00
RT-GRS	Jul	−14.4	−13.7	−15.2	0.00
RT-GRS	Aug	−11.7	−11.0	−12.4	0.00
RT-SS	Jun	−1.66	−1.22	−2.07	0.00
RT-SS	Jul	−1.17	−0.84	−1.49	0.002
RT-SS	Aug	−1.46	−1.46	−1.11	0.00
RT-RR	Jun	−7.47	−6.82	−8.15	0.00
RT-RR	Jul	−9.78	−9.45	−10.08	0.00
RT-RR	Aug	−8.05	−7.72	−8.39	0.00
RT-RMR	Jun	−6.14	−5.55	−6.75	0.00
RT-RMR	Jul	−8.69	−8.40	−9.0	0.00
RT-RMR	Aug	−8.29	−7.94	−8.65	0.00
RT-TR	Jun	−6.14	−5.54	−6.74	0.00
RT-TR	Jul	−10.96	−10.67	−11.24	0.00
RT-TR	Aug	−10.07	−9.74	−10.42	0.00
RT-WVH	Jun	−0.32	−0.26	−0.37	0.38
RT-WVH	Jul	−0.74	−0.67	−0.80	0.00
RT-WVH	Aug	−0.56	−0.48	−0.63	0.02

Thermal Refuge Area Occupancy Modeling

The GLMMs indicated that mean weekly river temperature provided the best overall fit for predicting Muskellunge presence in TRAs during the summer. This was supported by a BIC difference (ΔBIC) greater than 2.00 relative to the next closest model and a model weight of 0.70 (Table 3). Despite year-to-year variation in river flow data, neither mean daily flow nor mean monthly flow was found to accurately predict Muskellunge TRA occupancy (Table 3), suggesting a temperature-only effect. Additionally, neither sex nor TL was determined to improve model fit for Muskellunge occupancy in TRAs (Table 3). Our top model (weekly mean river temperature) indicated that the probability of TRA occupancy was approximately 10% at 21°C, 50% at 23–24°C, and over 90% at temperatures above 26°C (Figure 4). Dispersion parameters for all models were less than 1.1, and P-values for all paired fitted versus observed residuals were greater than 0.05, indicating good overall model fit.

TABLE 3. Summary of nine candidate generalized linear mixed models for predicting the presence of Muskellunge in thermal refuge areas (TRAs) during summer months (2017–2020). Model ranks are based on Bayesian information criterion (BIC) score and model weight (MRT = mean river temperature, °C; flow was in cubic meters per second; logL = log likelihood; ΔBIC = difference in BIC between the given model and the best-performing model).

Model	logL	BIC	ΔBIC	BIC weight
Weekly MRT	−164.80	347.80	0.00	0.70
Daily MRT	−166.30	350.80	3.00	0.16
Weekly MRT + TL	−163.90	352.00	4.20	0.09
Weekly MRT + sex	−164.70	353.50	5.70	0.04
Weekly MRT × log(weekly flow)	−163.10	356.50	8.70	0.01
Weekly MRT + sex + TL	−163.90	358.10	10.30	0.00
Daily temp × log(daily flow)	−164.30	358.80	11.00	0.00
Daily mean flow	−228.10	474.40	126.60	0.00
Weekly mean flow	−230.00	478.20	130.40	0.00

DISCUSSION

The results from this study suggest that seasonal movement patterns of Muskellunge are variable, with the greatest overall movement rates during the spring spawning season and the lowest movement rates during the summer period. These findings are consistent with other Muskellunge telemetry projects in southern water bodies, including a river (Brenden et al. 2006) and an impoundment (Morrison and Warren 2015). Most importantly, we documented high habitat affinity for TRAs during the summer. The greatest increase in TRA occupancy rates by Muskellunge in this study occurred between 23°C and 26°C, with the percentage of fish in TRAs increasing from approximately 30% to almost 90%. Above 27°C, Muskellunge were seen almost entirely in TRAs, with little variance.

Muskellunge appeared to select TRAs based on proximity rather than the magnitude of temperature difference. However, proximity versus temperature was not modeled because we did not have temperature loggers at all 11 TRAs. Despite direct groundwater influences (≤16°C) being available in each of the study areas, Muskellunge selected the closest available cooler-water area relative to their location. For example, fish 221 displayed the greatest movements over 3 years of tracking. This fish was initially tagged during the spawning run in 2018, moved back downstream over 12,000 m, and was detected in one location for a short time but then moved to the closest TRA for the summer. This pattern was repeated for 3 years. Not all Muskellunge displayed high movement rates like fish 221, but they still found the TRA that was closest to their summer ranges. Miller and Menzel (1986) documented Muskellunge exhibiting nonreproductive homing behavior throughout the summer months in an Iowa reservoir. Wagner and Wahl (2011) also reported decreased movements and a reduced home range during summer in an Illinois lake.

The reliance on thermal refugia by freshwater fishes during summer months is well documented (Gibson 1966; Moss 1985; Berman and Quinn 1991; Baird and Krueger 2003; Goniea et al. 2006). Bilby (1984) classified coldwater influences in a riverine system as cooler tributaries, seeps, deep pools, and cold alcoves, further defining thermal refugia as any plume with a temperature difference of at least 3°C less than ambient water temperature, a surface area of at least 0.5 m², and a DO level of at least 3 mg/L (Bilby 1984; Ebersole et al. 2003). Torgersen et al. (1995) determined that Chinook Salmon Oncorhynchus tshawytscha selected refuge habitats where water temperatures were 1–3°C cooler during thermally stressful periods. However, research documenting reliance on TRA use is limited for coolwater species and nonexistent for Muskellunge. To our knowledge, this is the first study to demonstrate the significance of TRAs during summer months for Muskellunge.

This study occurred in the Ridge and Valley physiographic province of Maryland, which contains the highest population densities of Muskellunge within this system (MDDNR 2017). Duigon (2001) determined that 89% of this area is of karst geology. Karst landscapes provide groundwater influences with consistent flow and temperature relief during summer months (Caldwell et al. 2020). The geological complexity within this system presented some challenges for capturing the magnitude of temperature differences between the refugia and main stem since several of the documented TRAs contained multiple small springs and seeps and did not have point-source-defined inflows. For example, the West Virginia Musky Hole, which was slightly cooler than the main stem (though not significantly so for June), contained multiple small seeps along approximately 200 m of shoreline, and fish concentrated in this area during periods of high water temperature while other tracked fish were occupying other identified TRAs. However, the actual temperature at the West Virginia Musky Hole was likely much lower than the main-stem temperature, but because of the difficulty in identifying an obvious point source of cold water, the difference in temperature relative to the main stem was not as great as the differences detected for the other TRAs.

Summer water temperatures in the upper main-stem Potomac River (31°C; maximum temperature observed) often reach and can exceed dangerous limits for Muskellunge (30°C; Scott and Crossman 1973). Warming water temperatures within the Potomac River watershed have been documented (Kaushal et al. 2010; Rice and Jastram 2015). Observations of Muskellunge concentrating near creek mouths, springs, seeps, and upwellings during summer months have been evident since the establishment of this population (J.E.M., personal observation). Throughout this study, tagged Muskellunge sought out the cooler water available during summer months as main-stem river temperatures approached 23°C, which is supported by similar findings from reservoir studies (Beck and Brooks 2003; Cole and Bettoli 2014). The results of this study suggest that Muskellunge rely heavily on TRAs when water temperatures exceed 27.0°C (>90% occupancy). During refuge occupancy, Muskellunge were extremely sedentary, were often visually observed in TRAs with layers of silt on them, and occasionally triggered the transmitter mortality sensors. Once main-stem river temperatures cooled, the fish resumed movements and deactivated the mortality signals. During 2018, one of the wettest summers on record, greater movements and less use of refuge habitats were observed compared to the summers of 2017, 2019, and 2020. Increased precipitation in 2018 corresponded to decreased main-stem temperatures and less attraction to refuge sites; however, this was not evident in our models because we were not interested in year-specific effects.

The lack of activity and the concentration of fish at TRAs suggest that Muskellunge are experiencing temperature-related physiological stress. Higher water temperatures can alter the blood oxygen affinity so that fish may not absorb adequate oxygen during periods of elevated temperatures even though oxygen is fully saturated in the water column (Helfman et al. 2009). Crane et al. (2020) reported false annuli in southern Muskellunge populations during summer months, implying that thermal stress was inhibiting growth. Under conditions of thermal stress, additional stress related to angling and handling may result in delayed release mortality. Landsman et al. (2011) reported significant increases in blood glucose, lactic acid accumulations, and plasma potassium concentrations from angled Muskellunge in warmer waters (maximum = temperature 26°C) on the Ottawa River, Canada, although no mortality was observed. Temperatures during the present study routinely exceeded 26°C, suggesting that stress levels for Muskellunge in the Potomac River are likely greater than those documented by Landsman et al. (2011) for fish in the Ottawa River. Beggs et al. (1980) concluded that these physiological effects can be responsible for mortality due to cardiac failure.

Perhaps unique to the freshwater Potomac River, many of these refugia are easily accessible to anglers due to the proximity and accessibility of the C&O Canal towpath. The towpath is a dirt-and-stone path that parallels the Potomac River for 297 km between Georgetown and Cumberland, Maryland. Sixteen (44%) of the 36 Muskellunge from this study were reported by anglers as being recaptured at least once. Six of those 16 Muskellunge were reported as being caught during the summer months. In total, 11 tagged Muskellunge died during this study: 6 in the spring period, and 5 in the summer. Only one mortality during the summer period was directly attributed to angling, whereas the other four fates are unknown. Fish 870 was angled from a thermal refuge during the summer 2017. This fish, a 1,067-mm female, was harvested after an unsuccessful release attempt. All other angled fish were collected outside of the summer period and survived for over 1 month postcapture, suggesting that there are few adverse effects associated with angling by experienced anglers in nonsummer months.

Additionally, angled Potomac River Muskellunge reported through a voluntary creel diary program indicated that catch rates are highest during July when fish are initially confined to thermal refugia, but they decline in August, potentially after prolonged exposure to thermal stress (MDDNR 2019). Angler tag returns from an ongoing tagging program (2000–2019) were also highest during the summer months, and an angler preference survey indicated that summer was the most popular season for Muskellunge angling in the Potomac River (MDDNR 2017). The data demonstrate that anglers successfully target and catch Muskellunge during the summer period and when Muskellunge occupy TRAs. Furthermore, results of a 2020 angler response survey that was specific to the freshwater Potomac River showed an extremely low response rate of 21% (Maryland Department of Natural Resources, unpublished data), suggesting that more radio-tagged fish were likely caught during the summer but were not reported.

This study provides direct evidence of Muskellunge reliance on/occupancy of TRAs during periods of thermal stress in a southern riverine population. Furthermore, the temperature occupancy model demonstrates a need to better understand summer angling mortality, especially given the projected rise in temperatures in the decades to come. Managers of southern riverine Muskellunge fisheries should consider identifying and protecting TRA habitats in main-stem rivers where temperatures exceed 25°C for extended periods of time during summer months. It is quite possible that Muskellunge population size, growth, and mortality may be limited to TRA availability in southern rivers. This additional information will help managers to determine whether outreach or regulatory actions are warranted to sustain these trophy fisheries if summer angling pressure and/or exploitation is high.